_zoltan_

6 ms·

91.
by _zoltan_ 4mo ago
they exist, sure. And I'm sure it can handle PB+. on prem is an existing market, however, if you reread my comment I talked about running minio on AWS because S3 is too expensive - just doesn't make sense to do. I've yet to m
92.
by _zoltan_ 4mo ago
up to a limited number of snapshots. it's not arbitrary times, it's when there was an atomic snapshot created in the catalog.
93.
by _zoltan_ 4mo ago
your reply makes zero sense. there is a reason why people develop for S3: a lot of enterprise data is there. people ingest there from various sources. and it's not just parquet usually, it's multivendor sources writing to an icebe
94.
by _zoltan_ 4mo ago
I disagree. I think it's very important to see how can those papers be turned into real things. Rejecting something for too practical is a joke, and if I ever find out who gave us that review, I'll have a talk with them.
95.
by _zoltan_ 4mo ago
I can e-mail wise now with my bill and pay it. it's AWESOME. I've used a bill that was in Hungarian for a EUR account in Austria. No issue. My swiss banks won't be able to do this in the next 5 years. I am not affiliated with
96.
by _zoltan_ 4mo ago
that's a very condescending view from you, to be honest.
97.
by _zoltan_ 4mo ago
our paper to a database venue about bringing GPU support to Presto was rejected. one of the reviewers wrote, and I quote verbatim: "the topic of the paper is too practical". I just couldn't help but laughed at it.
98.
by _zoltan_ 4mo ago
I just wanted to commend you on this analogy. Very nice. Thank you!
99.
by _zoltan_ 4mo ago
it's not stalled, as that would imply that it waits for something, which is not necessarily the case with bubbles. most often it shows lack of proper pipelining or wrong pipeline dependencies (pipeline A waits for pipeline B, pipeline
100.
by _zoltan_ 4mo ago
while the title is misreading, when reading GPU profiling data, we do call these bubbles - where the GPU _could_ do something, but it's idle. any time your GPU is idle = you are losing $$$ = your TCO is going up. you don't want th
101.
by _zoltan_ 4mo ago
there is a difference but it's overrated. if a theorem is proven, then, as OP said, the theorem is the interface, no matter where the proof is. just as we don't re-prove Fermat's little theorem every time I use it in a proo
102.
by _zoltan_ 4mo ago
I gave them money once on Kickstarter. Never got anything and they held my money hostage for years. I'm never ever, ever buying anything from Jolla. They can go out of business for all I care.
103.
by _zoltan_ 4mo ago
ever heard of RoCE? what latency are you thinking of in the Spectrum-X 800Gbit RoCE era?
104.
by _zoltan_ 4mo ago
like?
105.
by _zoltan_ 4mo ago
in 2026 you can full the PCIe BW from remote storage at roughly the same latency as your local NVMe drive (hello NVMeoF). your comment would have been valid 10 years ago, maybe.
106.
by _zoltan_ 4mo ago
also living in Swizerland and I disagree. Hard. it's horrible that Europe is so backwards in AI. too much regulation and nothing to show for it. we should be way faster. there is no money. the culture in both Europe and Switzerland is
107.
by _zoltan_ 4mo ago
for corporate use it's a mistake not to use a frontier model.
108.
by _zoltan_ 4mo ago
DHH is even less known, don't kid yourself.
109.
by _zoltan_ 4mo ago
no, most people wouldn't know. you're in an echo chamber if you think he is well known.
110.
by _zoltan_ 4mo ago
We're discussing home use. You can. You just don't want to. Huge difference.
111.
by _zoltan_ 4mo ago
that's not the LLM's fault. I iterate and give very specific instructions and my code is really good.
112.
by _zoltan_ 4mo ago
your two statements can be true at the same time: I do not need to look at the code because I know exactly what it does. do I care for every snippet? every call's signature? no I do not. do I understand what it does 100%? yes, because
113.
by _zoltan_ 4mo ago
The code is not meant to be read anymore. That's the mistake. Do you look at generated assembly that comes out of your compiler? You don't. So why are you looking at this code? We have pushed up the abstraction layer.
114.
by _zoltan_ 4mo ago
20 dollars for this is practically free.
115.
by _zoltan_ 4mo ago
I've been doing the same. take papers, define a high level goal, then let it iterate. I have access to DGX boxes and watching the model rewrite stuff to take NVLink into account after it discovered it was great :-)
116.
by _zoltan_ 5mo ago
I don't know if you guys are trolling or not, honestly. Vibe coding is misrepresented. I'm building extremely complex features that are grounded in our codebase but fundamentally rests on the training data which includes all acade
117.
by _zoltan_ 5mo ago
I am not aware of any commercial vendor shipping v3/v4 boards with DDR3. I have a couple hundred Supermicro systems that are stuck on v2 CPUs with DDR3...
118.
by _zoltan_ 5mo ago
I can't give exact numbers but I find the cost more than OK.
119.
by _zoltan_ 5mo ago
compared to salaries, our bill isn't outrageous.
120.
by _zoltan_ 5mo ago
right, and they talk about "v4" which is DDR4.

More ›